Class | Needle::Container |
In: |
lib/needle/container.rb
|
Parent: | Object |
The container is the heart of Needle’s model. Every Container instance is a miniature registry, and is really a namespace separate from every other Container instance. Service lookups inside of a container always look in self first, and if not found, they then look in their parent container, recursively.
You will rarely need to instantiate a Container directly. Instead, use the Container#namespace method to create new containers.
defaults | [R] | A hash of default options to use when registering services. These defaults also apply to namespaces, so when specifying a new default service model (for instance) there may be unexpected side-effects with the namespaces that are created. |
name | [R] | The name of this container. May be nil. |
parent | [R] | The container that contains this container. This will be nil for the root of a hierarchy (see Registry). |
Create a new empty container with the given parent and name. If a parent is given, this container will inherit the defaults of the parent at the time the container was created.
If a block is given, yields the container’s builder instance to the block. Otherwise, simply returns the builder instance.
Usage:
container.define do |b| b.foo { Bar.new } b.baz { Baz.new } ... end
Or:
container.define.foo { Bar.new } container.define.baz { Baz.new }
Create a new DefinitionContext around the container, and then evaluate the block within the new context instance (via instance_eval).
Usage:
container.define! do calc( :model => :prototype ) { Calc.new( operations ) } end
Returns true if this container either is the given container or is descended from the given container, and false otherwise.
Searches the current container and its ancestors for the named service. If found, the service point (the definition of that service) is returned, otherwise nil is returned.
Return the fully qualified name of this container, which is the container’s name and all parent’s names up to the root container, catenated together with dot characters, i.e., "one.two.three".
Retrieves the named service, if it exists. Ancestors are searched if the service is not defined by the current container (see find_definition). If the named service does not exist, ServiceNotFound is raised.
Note that this returns the instantiated service, not the service point.
Also, if any pipeline element in the instantiation pipeline does not support extra parameters when extra parameters have been given, then an error will be raised.
Returns true if this container includes a service point with the given name. Returns false otherwise.
Describe a new interceptor to use that will intercept method calls on the named service. This method returns a new Interceptor instance, which can be used directly to configure the behavior of the interceptor.
Usage:
container.intercept( :calc ).with { |c| c.logging_interceptor }
Returns true if this container or any ancestor includes a service point with the given name. Returns false otherwise.
As a convenience for accessing services, this delegates any message sent to the container (which has no parameters and no block) to Container#[]. Note that this incurs slightly more overhead than simply calling Container#[] directly, so if performance is an issue, you should avoid this approach.
Usage:
container.register( :add ) { Adder.new } p container.add == container[:add] # => true
Create a new namespace within the container, with the given name. If a block is provided, it will be invoked when the namespace is created, with the new namespace passed to it.
For the curious, namespaces are simply services that are implemented by Container. The two statements are conceptually identical:
container.namespace( :calc ) container.register( :calc ) { |c,p| Needle::Container.new( c, p.name ) }
Note that this means that namespaces may be singletons or prototypes, or have immediate or deferred instantiation, and so forth. (The default of immediate, singleton instantiation is sufficient for 99% of the things you’ll use namespaces for.)
Usage:
container.namespace( :operations ) do |op| op.register( :add ) { Adder.new } ... end adder = container.calc.operations.add
Note: the block is not invoked until the namespace is created, which is not until it is first referenced. If you need the namespace to be created immediately, either use namespace_define or reference the namespace as soon as you’ve created it.
Create a new namespace within the container, with the given name. The block (which is required) will be passed to Container#define on the new namespace.
For the curious, namespaces are simply services that are implemented by Container. The two statements are really identical:
container.namespace( :calc ) container.register( :calc ) { |c,p| Needle::Container.new( c, p.name ) }
Note that this means that namespaces may be singletons or prototypes, or have immediate or deferred instantiation, and so forth. (The default of immediate, singleton instantiation is sufficient for 99% of the things you’ll use namespaces for.)
Usage:
container.namespace_define( :operations ) do |b| b.add { Adder.new } ... end adder = container.calc.operations.add
Note: this method will immediately instantiate the new namespace, unlike namespace. If you want instantiation of the namespace to be deferred, either use a deferring service model (like :singleton_deferred) or create the namespace via namespace.
Create a new namespace within the container, with the given name. The block (which is required) will be passed to Container#define! on the new namespace.
For the curious, namespaces are simply services that are implemented by Container. The two statements are really identical:
container.namespace( :calc ) container.register( :calc ) { |c,p| Needle::Container.new( c, p.name ) }
Note that this means that namespaces may be singletons or prototypes, or have immediate or deferred instantiation, and so forth. (The default of immediate, singleton instantiation is sufficient for 99% of the things you’ll use namespaces for.)
Usage:
container.namespace_define!( :operations ) do add { Adder.new } ... end adder = container.calc.operations.add
Note: this method will immediately instantiate the new namespace, unlike namespace. If you want instantiation of the namespace to be deferred, either use a deferring service model (like :singleton_deferred) or create the namespace via namespace.
Returns the pipeline object for the named service, which allows clients to explicitly manipulate the service’s instantiation pipeline.
Usage:
container.pipeline( :calc ). add( :initialize ). add( :custom ) { |me,*args| me.succ.call( *args ) }
Register the named service with the container. When the service is requested (with Container#[]), the associated callback will be used to construct it.
This returns the registry that was used to register the service.
Usage:
container.register( :calc, :model=>:prototype ) do |c| Calc.new( c.operations ) end
Require the given file, and then invoke the given registration method on the target module. The container will be passed as the sole parameter to the registration method. This allows you to easily decentralize the definition of services.
Usage:
container.require( "app/services", "App::Services" ) # in app/services.rb: module App module Services def register_services( container ) ... end module_function :register_services end end
Returns true if this container responds to the given message, or if it explicitly contains a service with the given name (see has_key?). In this case, has_key? is used instead of knows_key? so that subcontainers may be used as proper hashes by their parents.
Returns the root of the current hierarchy. If the container is the root, returns self, otherwise calls Container#root on its parent. The value is cached for future reference.
Specifies a set of default options to use temporarily. The options are merged with the current set of defaults for the container. The original options are returned, and may be restored by invoking use again with the hash that is returned. If a block is given, the registry will be yielded to it and the options automatically restored when the block returns.
Specifies a set of default options to use temporarily. The original options are returned. This differs from use in that it will completely replace the original options, instead of merging the parameters with the originals.